We investigate the effects of potential sources of systematic error on the angular and photometric redshift, z_phot, distributions of a sample of redshift 0.4 < z < 0.7 massive galaxies whose selection matches that of the Baryon Oscillation Spectroscopic Survey (BOSS) constant mass sample. Utilizing over 112,778 BOSS spectra as a training sample, we produce a photometric redshift catalog for the galaxies in the SDSS DR8 imaging area that, after masking, covers nearly one quarter of the sky (9,913 square degrees). We investigate fluctuations in the number density of objects in this sample as a function of Galactic extinction, seeing, stellar density, sky background, airmass, photometric offset, and North/South Galactic hemisphere. We find that the presence of stars of comparable magnitudes to our galaxies (which are not traditionally masked) effectively remove area. Failing to correct for such stars can produce systematic errors on the measured angular auto-correlation function, w, that are larger than its statistical uncertainty. We describe how one can effectively mask for the presence of the stars, without removing any galaxies from the sample, and minimize the systematic error. Additionally, we apply two separate methods that can be used to correct the systematic errors imparted by any parameter that can be turned into a map on the sky. We find that failing to properly account for varying sky background introduces a systematic error on w. We measure w, in four z_phot slices of width 0.05 between 0.45 < z_phot < 0.65 and find that the measurements, after correcting for the systematic effects of stars and sky background, are generally consistent with a generic LambdaCDM model, at scales up to 60 degrees. At scales greater than 3 degrees and z_phot > 0.5, the magnitude of the corrections we apply are greater than the statistical uncertainty in w.